Target Identification Technology in Space Light Environment Based on Improved YOLOX

The space light environment is complicated, and the target’s reflection characteristics are very different from the ground as so as the target’s contrast is large and it is prone to overexposure. In view of the shortage of the existing target identification technology in space light environment under traditional way, this thesis puts forward a model of target identification based on improved YOLOX in the space lighting environment, and according to the characteristics of the space target image, this thesis designed a special loss function for space image optimization. It realizes real-time detection of target, improving the accuracy of the target detection. Aiming at the special space lighting environment, the dataset of space target image is constructed. The dataset simulates the real space environment, and the lighting has serious problems such as strong reflection, blurring, angle tilt, fragmentation, occlusion, etc. After the optimization training of this dataset based on the improved YOLOX network, the accuracy of target detection has been significantly improved. Experimental results show that the improved algorithm in this paper improves 27.41% compared with the traditional algorithm and 11.12% compared with the original YOLOX algorithm, which can effectively identify targets in the space complex lighting environment.